This paper presents a novel approach to the personalization of natural language dialogue agents. The adaptation to a given user relies on usage modeling at the interaction level. The purpose of this paper is to apply such an approach to adaptation within the dialogue agent (i.e. the interface). The goal is to make the dialogue agent behave proactively. Our approach is characterized by its genericity: it is defined independently from the types of speech acts, from the mode of interaction and from the application behind the dialogue agent. We show how such an approach can be relevant in the case of natural language dialogue and how it contributes to user's acceptance of the related supported services.